k8ika0s commented on PR #48221: URL: https://github.com/apache/arrow/pull/48221#issuecomment-3568382232
@Vishwanatha-HD Thanks for taking a look at the buffer and OOM paths — these tests always feel a little fragile, and every platform seems to expose a different edge case. One thing I’ve run into on s390x is that some of the allocators (mimalloc in particular) don’t fail gracefully on the “impossible size” cases. They sometimes hit fatal paths before Arrow ever gets a chance to raise `OutOfMemory`, which can make the test suite look broken even though the actual Arrow logic is fine. I noticed this version skips the test entirely on big-endian builds. That definitely avoids the allocator crash, though it does mean BE never exercises the OOM branch at all. In my own testing I’ve had some luck clamping the huge allocation down to something like `1 << 48` and only skipping when the allocator backend is known to abort, which keeps the OOM behavior testable without triggering allocator issues. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
